3 Key Takeaways
Plain English Summary- 01Core Finding
We used unsupervised machine learning based on GADA status, fasting C-peptide, glucose, triglycerides, and HDL cholesterol to characterise clinical and metabolic type 1 diabetes subtypes.
- 02Study Evidence
Tested in a genetic / mendelian study + observational cohort study (n=600) at the real-world cohort stage.
- 03Practical Meaning
Genetic / Mendelian randomization findings highlight causal biological risk, but ancestry coverage and pleiotropy can limit generalizability.
What This Means for You
Our results indicate that individuals with combined autoimmunity and lower insulin sensitivity present with clinical diabetes earlier than classic type 1 diabetes along the β-cell decline trajectory, identifying a potential timeframe for disease-modifying interventions in specific type 1 diabetes subtypes.
The Discovery
Adult-onset type 1 diabetes exhibits substantial heterogeneity. We aimed to investigate whether this variability reflects distinct disease entities or different temporal stages of a shared autoimmune process. In this prospective cohort study, we analysed data from adults (aged 18-69 years) in the German Diabetes Study who had clinically diagnosed type 1 diabetes within the preceding 12 months, and from individuals who were positive for glutamic acid decarboxylase antibody (GADA) but had originally been diagnosed with type 2 diabetes. We used unsupervised machine learning based on GADA status, fasting C-peptide, glucose, triglycerides, and HDL cholesterol to characterise clinical and metabolic type 1 diabetes subtypes. The resulting subtypes were compared for aetiological traits (polygenic risk scores for type 1 diabetes and type 2 diabetes, partitioned type 2 diabetes polygenic risk scores, human leukocyte antigen [HLA] profiles, and islet-directed autoantibodies), insulin sensitivity, glucagon-stimulated C-peptide trajectories as a measure of residual β-cell function, and complications during prospective follow-up. Reproducibility was assessed in adults with type 1 diabetes from the German Diabetes Prospective Follow-up Registry (Diabetes-Patienten-Verlaufsdokumentation [DPV]). Analysed participants in the German Diabetes Study cohort were recruited between June 27, 2005, and Jan 20, 2025 (n=600), and were followed up for a mean of 6·1 (SD 4·7) years. We identified three type 1 diabetes subtypes: two GADA-positive subtypes, and one GADA-negative subtype. A similar clinical and metabolic subtype structure was observed in the DPV cohort, whose participants were recruited between 2000 and 2025 (n=483) and were followed up for a mean of 3·1 (SD 5·0) years. One GADA-positive subtype was characterised by lower BMI, higher insulin sensitivity, and severe β-cell dysfunction (classic type 1 diabetes), whereas the second showed higher BMI, lower insulin sensitivity, and better β-cell function (mixed type 1 diabetes). The two GADA-positive subtypes did not differ in autoimmunity or type 1 diabetes polygenic risk scores, whereas the GADA-negative subtype had lower genetic type 1 diabetes risk, despite no clear subtype-specific differences in HLA risk haplotypes. Longitudinal modelling indicated that the mixed type 1 diabetes group was positioned 6·1 years (95% CI 4·6-7·8) earlier along the β-cell function decline trajectory compared with the classic type 1 diabetes group. The risks of hypoglycaemia (log-rank p=0·0019), retinopathy (mixed type 1 diabetes vs classic type 1 diabetes: adjusted hazard ratio 4·93 [95% CI 1·71-14·24]), and cardiovascular autonomic neuropathy (log-rank p=0·028) differed across the groups.
How It Works, Simply
This endocrinology research focuses on LDL cholesterol · Antibodies · Insulin. Researchers evaluated it using a genetic / Mendelian study + Observational cohort study (n=600) at the real-world cohort stage of medical research.
3 Key Takeaways
Clinical Takeaways- 01Core Finding
We used unsupervised machine learning based on GADA status, fasting C-peptide, glucose, triglycerides, and HDL cholesterol to characterise clinical and metabolic type 1 diabetes subtypes.
- 02Study Evidence
Tested in a genetic / mendelian study + observational cohort study (n=600) at the real-world cohort stage.
- 03Practical Meaning
Genetic / Mendelian randomization findings highlight causal biological risk, but ancestry coverage and pleiotropy can limit generalizability.
Biological Mechanism & Study Design
Study Design & Cohort
Genetic / Mendelian study + Observational cohort study
Genetic / Mendelian study + Observational cohort study (n=600) · Real-World Cohort
Biological Pathway & Modality
LDL cholesterol · Antibodies · Insulin
Intervention: Clinical and metabolic subtypes of recent-onset adult type 1 diabetes: a prospective
Experimental Findings & Endpoints
Adult-onset type 1 diabetes exhibits substantial heterogeneity.
We aimed to investigate whether this variability reflects distinct disease entities or different temporal stages of a shared autoimmune process. In this prospective cohort study, we analysed data from adults (aged 18-69 years) in the German Diabetes Study who had clinically diagnosed type 1 diabetes within the preceding 12 months, and from individuals who were positive for glutamic acid decarboxylase antibody (GADA) but had originally been diagnosed with type 2 diabetes.
We used unsupervised machine learning based on GADA status, fasting C-peptide, glucose, triglycerides, and HDL cholesterol to characterise clinical and metabolic type 1 diabetes subtypes. The resulting subtypes were compared for aetiological traits (polygenic risk scores for type 1 diabetes and type 2 diabetes, partitioned type 2 diabetes polygenic risk scores, human leukocyte antigen [HLA] profiles, and islet-directed autoantibodies), insulin sensitivity, glucagon-stimulated C-peptide trajectories as a measure of residual β-cell function, and complications during prospective follow-up.
Reproducibility was assessed in adults with type 1 diabetes from the German Diabetes Prospective Follow-up Registry (Diabetes-Patienten-Verlaufsdokumentation [DPV]). Analysed participants in the German Diabetes Study cohort were recruited between June 27, 2005, and Jan 20, 2025 (n=600), and were followed up for a mean of 6·1 (SD 4·7) years.
We identified three type 1 diabetes subtypes: two GADA-positive subtypes, and one GADA-negative subtype. A similar clinical and metabolic subtype structure was observed in the DPV cohort, whose participants were recruited between 2000 and 2025 (n=483) and were followed up for a mean of 3·1 (SD 5·0) years.
One GADA-positive subtype was characterised by lower BMI, higher insulin sensitivity, and severe β-cell dysfunction (classic type 1 diabetes), whereas the second showed higher BMI, lower insulin sensitivity, and better β-cell function (mixed type 1 diabetes). The two GADA-positive subtypes did not differ in autoimmunity or type 1 diabetes polygenic risk scores, whereas the GADA-negative subtype had lower genetic type 1 diabetes risk, despite no clear subtype-specific differences in HLA risk haplotypes.
Longitudinal modelling indicated that the mixed type 1 diabetes group was positioned 6·1 years (95% CI 4·6-7·8) earlier along the β-cell function decline trajectory compared with the classic type 1 diabetes group. The risks of hypoglycaemia (log-rank p=0·0019), retinopathy (mixed type 1 diabetes vs classic type 1 diabetes: adjusted hazard ratio 4·93 [95% CI 1·71-14·24]), and cardiovascular autonomic neuropathy (log-rank p=0·028) differed across the groups.
Clinical Translation & Practice Context
Our results indicate that individuals with combined autoimmunity and lower insulin sensitivity present with clinical diabetes earlier than classic type 1 diabetes along the β-cell decline trajectory, identifying a potential timeframe for disease-modifying interventions in specific type 1 diabetes subtypes.
The Bottom Line · Clinical Pearl
“Our results indicate that individuals with combined autoimmunity and lower insulin sensitivity present with clinical diabetes earlier than classic type 1 diabetes along the β-cell decline trajectory, identifying a potential timeframe for disease-modifying interventions in specific type 1 diabetes subtypes.”
Published in The Lancet Diabetes & Endocrinology · Current Stage: Real-World Cohort